How we test, rate and audit claims

Last updated 7 September 2026

Most product sites blur two very different things: whether the reviewer physically used a product, and whether the manufacturer's claims about it hold up. We keep them on separate axes, because a product can be thoroughly tested and still make unsupported claims — and a claim can be well supported by research nobody on our team has ever handled the device to verify.

Axis one — what we actually did with the product

Every product on a commercial page carries exactly one of these, and it is shown on the page itself, not buried here.

Label	What it means	What it does not mean
Researched	We compared published specifications, manufacturer documentation, pricing, warranty terms and independent sources.	We did not physically use the product. We will never imply that we did.
Hands-On	We physically had the product and evaluated it, and we state exactly what we did and for how long.	Not a laboratory measurement unless the page says a measurement was taken and names the instrument.
Long-Term Tested	Extended documented use, with the duration and conditions disclosed on the page.	Not a clinical trial, and not a claim about outcomes for anyone but us.

Axis two — whether the claims hold up

Claim Audited is a separate status. It means we took the specific claims a manufacturer makes and tried to trace each one back to the research it rests on. A product can be Claim Audited without our ever having touched it, and it can be Long-Term Tested without having been audited. That is why they are two labels and not a single score.

Each audited claim ends in one of three places:

Traced to human research — we found published research in humans that the claim appears to rest on, and we cite it.
Traced to weaker evidence — the support exists but is animal, in vitro, or otherwise not a human outcome, and we say which.
Could not substantiate — we could not find the support. This means we could not find it. It is not a finding that the claim is false.

The evidence hierarchy we use

Not all support is equal, and we rank it in this order:

Systematic reviews and meta-analyses of human trials.
Randomised controlled trials in humans.
Other human studies — observational, uncontrolled, or very small.
Animal or laboratory work.
Manufacturer testing that has not been independently published.
Testimonial and marketing copy, which is not evidence and is never cited as though it were.

Manufacturer claims are kept separate from findings

A specification we have taken from a manufacturer is labelled as the manufacturer's figure. A number we measured is labelled as measured. These are never presented in the same column as if they were equivalent, because for the category we are starting with they routinely are not.

Why measurement distance matters, and where most comparisons go wrong

Our launch cluster is red light therapy panels, where the headline number is irradiance — how much light reaches you. That figure is meaningless without the distance it was measured at, and manufacturers quote it at different distances. A panel that looks twice as powerful may simply have been measured half as far away.

So we record the distance alongside every irradiance figure, and we will not publish a comparison of irradiance values measured at distances we cannot establish. If we cannot get comparable numbers, we say the comparison cannot be made honestly rather than printing a table that looks authoritative and is not.

Pricing and freshness

Prices change and links rot. Every commercial page records when its prices were last verified, and we re-check on a schedule rather than leaving a two-year-old figure on the page. If a product is discontinued or a link stops resolving to the product it claims to, we remove or correct it.

Money

Some links earn us a commission, disclosed automatically above the first such link on any page that has one. Compensation plays no part in ranking, scoring, or whether a claim is recorded as substantiated. The full position is on the Affiliate Disclosure page, and our wider standards are on the Editorial Policy page.

When we get it wrong

We will. When it is pointed out, we correct it and say what changed — see Corrections.